Protocol 3 - Geochemical and petrographic data
The following example applies protocol 3 to confirm workshops’ provenance groups.
Protocol 3 consist in:
- Select geochemical compositional data (CoDa) and ordinal petrographic data;
- Centred log-ratio transformation (clr) and transform to ranks;
- Extended Gower distance, using Relative ranking difference (RRD);
- Apply Principal Coordinates Analysis (PCoA);
- Perform PERMANOVA & PERMDISP tests;
Last, search for outliers and re-do protocol excluding outliers.
NOTE: The initial procedures must be ran at least once before any protocol can be applied.
See protocol 2, for consulting references on the extended Gower distance.
Ordination procedure
Protocol 3 performs PCoA on a distance matrix calculated with Extended Gower coefficient of dissimilarity, combining Euclidean distances on transformed compositional data (50%) and RRD on ranked petrographic data (50%). As in protocol 2, PCoA requires specifying the number of dimensions and so you must 2D and 3D ordination objects separately:
prot3_2d <- apply_ordination(cleanAmphorae[!isShipwreck,], # no shipwrecks
"3", # select protocol 3
exception_columns = excep_cols,
variable_tags = varCode,
coda_override = chemVars16,
coda_transformation = "CLR")
prot3_3d <- apply_ordination(cleanAmphorae[!isShipwreck,], # no shipwrecks
"3", # select protocol 3
exception_columns = excep_cols,
variable_tags = varCode,
coda_override = chemVars16,
coda_transformation = "CLR",
dimensions = 3)Simplify CoDa names
We may want to simplify the names of the transformed variables before plotting them in a biplot.
prot3_2d <- simplify_coda_names(prot3_2d)
prot3_3d <- simplify_coda_names(prot3_3d)Test the given provenance groups
Because protocol 3 uses both geochemical and petrographic information, we can test the provenance assigned to the amphorae samples.
prot3_tests <- test_groups(prot3_2d$dist_matrix,
factor_list$ProvGroup)
#> [1] "initiating test batch..."
#> [1] "vegan::anosim done."
#> [1] "vegan::betadisper done."
#> [1] "vegan::permutest done."
#> [1] "vegan::adonis done."
#> [1] "Test batch completed."These tests were explained in protocol 1.
Biplots
The details on how to create biplots is already explained in protocol 1. Unlike protocol 2, protocol 3 only generates one kind of projection (RRD, PCoA).
Biplot 2D
arrows_label_adj <- rbind(c(.5,.8),c(.5,1),c(.5,1),c(.5,0),c(.5,1),
c(.5,0),c(0,.5))
row.names(arrows_label_adj) <- c("L48","L24","L5","L36","S7",
"S8","S11")
biplot2d3d::biplot_2d(prot3_2d,
ordination_method = "PCoA",
invert_coordinates = c (TRUE,TRUE),
ylim = c(-.3,.29),
point_type = "point",
groups = factor_list$FabricGroup,
group_color = color_list$FabricGroup,
group_label_cex = 0.6,
arrow_mim_dist = 0.5,
arrow_label_cex = 0.6,
arrow_fig = c(.6,.95,0,.35),
arrow_label_adj_override = arrows_label_adj,
subtitle = prot3_2d$sub2D,
test_text = prot3_tests$text(prot3_tests),
test_fig = c(0, 0.5, 0.62, .99),
fitAnalysis_fig = c(0,.7,.05,.5),
output_type = "preview")
protocol 3
Biplot 3D
biplot2d3d::biplot_3d(prot3_3d,
ordination_method = "PCoA",
point_type = "point",
groups = factor_list$FabricGroup,
group_color = color_list$FabricGroup,
group_representation = "stars",
star_centroid_radius = 0,
star_label_cex = .8,
arrow_min_dist = .5,
arrow_body_length = .025,
subtitle = prot3_3d$sub3D,
test_text = prot3_tests$text(prot3_tests),
test_cex = 1.25,
test_fig = c(0, 0.5, 0.65, .99),
view_zoom = 0.9)
biplot2d3d::animation(directory = directories$prot3,
file_name = "Prot3_Biplot3D")
Prot3_Biplot3D.gif